This method should allow the created overlay network to, optionally, be attachable.

Docker::Swarm::Swarm.create_network_overlay crashes with Socket EOF exception when one (or more) of the nodes is down. When attempting to retrieve node.networks from a node that's currently down it returns a Socket EOF

In docker-api I could do:

Docker::Event.stream { |event| puts event }

It would be great to have the same functionality but with Swarm Events.

Thanks in advance!

Hello,

Is this repository/code not maintained anymore?

I noticed that quite a few issues or PRs are unanswered since a few years...

Thanks for your answer in advance.

Right now the tasks instance method (http://www.rubydoc.info/gems/docker-swarm-api/Docker/Swarm/Swarm#tasks-instance_method) of the Swarm class does not take any parameters and as such returns every tasks in a swarm.

It would be a good idea to add a parameter to the tasks method in order to filter by any valid filter criteria (e.g. service=<service name>) as defined in the Docker API here: https://docs.docker.com/engine/api/v1.27/#tag/Task

Using the following code:

swarm.nodes.each do |node|
  p node.networks
end

It will fail with the following error:

RuntimeError at ... No connection set for node: ...
lib/docker/swarm/node.rb in networks
148. raise "No connection set for node: #{self.host_name}, ID: #{self.id}"

% bundle exec rspec
    Sorry, you can't use byebug without Readline. To solve this, you need to
    rebuild Ruby with Readline support. If using Ubuntu, try `sudo apt-get
    install libreadline-dev` and then reinstall your Ruby.

Readline is installed.

The README.md speaks of create_overlay_network, perhaps swarm. create_network_overlay was meant, instead.

Initialize Swarm

def self.init(opts, connection)
...
swarm = Docker::Swarm::Swarm.swarm(opts, connection)
...
end

docker swarm join-token -q worker

def self.swarm(connection, options = {})
...

% bundle exec rake -T
rake aborted!
LoadError: cannot load such file -- cane/rake_task

Hi,

I discovered after after some fight with the docker api, that in order to actually force the update of a service, the ForceUpdate flag must actually change, not only be set to 1.

They describe it as a counter, not as a flag. So the current implementation of restart can only be used once, after that the ForceUpdate counter in the service spec (as reported by docker service inspect will be set to 1 and the restart won't work.